Thought disorder is a cardinal symptom of schizophrenia, and is inferred through abnormal language. The underlying cognitive abnormality that causes thought disorder is unknown. The broad aim of this proposal is to understand the neurophysiological basis of thought disorder in schizophrenia and psychotic mania. We will test a model of semantic over-activation and verbal Working memory dysfunction in schizophrenia, contrasted with an executive attention abnormality in mania. We will measure language comprehension, reaction times, and the activity of the brain via event-related potentials, and correlate these measures with clinical measures of symptoms and neuropsychological measures of working memory performance. One possible dysfunction is an abnormality in semantic memory activation (hyperpriming) whereby associated by contextually inappropriate concepts crowd the mind of schizophrenia patients. Paradoxically, schizophrenia patients outperform controls on speeded lexical decision tasks, with faster reaction times. However, at slower speeds, the performance of patients falls off sharply. The neural substrates of this hyperpriming and of the later drop off in performance are unknown. We test a two-factor model of schizophrenic thought disorder with both an initial automatic excitatory abnormality and a later abnormality in the controlled utilization of context stored in verbal working memory by having subjects process sentences, word pairs, and word triplets for semantic relatedness. The stimuli will include homographs as the primary probe of associative derangement. Homographs are words with multiple meanings, and occupy places in more than one unrelated semantic network. For example, board may mean plank or committee. Schizophrenia patients appear to display a semantic bias towards strong associates and networks such that they tend to select the dominant homograph meaning (board means plank) even in the face of context that mandates accessing the subordinate meaning (board means committee). We propose to examine language processing while varying networks strength (dominant versus subordinate meanings), while varying associative strength (strongly versus weakly associated words), in the face of strongly biasing context followed by weak associates, and in the face of weakly biasing context followed by strong associates. These different manipulations will be presented under fast (250 msec) and slow (1.25 sec) rates, to more heavily stress the semantic activation system and the verbal working memory system respectively. The use of ERPs allows for direct measures of stimulus processing, and provides some information about putative sources of brain dysfunction (when informed from previous direct recordings). These studies have the potential to help clarify the nature of cognitive dysfunction in schizophrenia and in psychotic mania.